Quick Answer
ChatGPT traffic does convert better. According to a 2026 study by Visibility Labs, based on 94 ecommerce brands, ChatGPT traffic converted at 1.81% against 1.39% for non-branded organic, which is the 31% gap everyone quotes. According to the same study, that traffic produced 1.48% of the revenue non-branded organic produced. Treat the conversion rate as evidence that AI users arrive further along in their decision, and treat the volume as the reason not to move budget out of channels that are currently paying your bills.
Put The Denominator Back
The study, reported by Search Engine Land in February 2026, covered 12 months of GA4 data from 94 seven- and eight-figure ecommerce brands. It compared 9.46 million non-branded organic sessions against 135,000 ChatGPT referral sessions.
Look at those two session counts again. Non-branded organic was roughly 70 times larger across the year, narrowing to 47 times larger by the fourth quarter.
The conversion rate is the number that travels. The session count is the number that decides your budget. Per that session split, a channel converting 31% better on roughly 1.4% of the volume is not competing with your organic programme. It is a small, early signal sitting next to it.
There is a further caution inside the data. Early in the period, some months carried only 15 to 37 ChatGPT-attributed conversions. Rates built on numbers that small move sharply on very little. The authors were clear that statistical confidence improved only around midyear.
What The Full Dataset Actually Says
All figures below are from the same Visibility Labs 2026 study.
| Measure | ChatGPT referral | Non-branded organic | What it tells you |
|---|---|---|---|
| Conversion rate | 1.81% | 1.39% | Buyers arrive more decided |
| Average order value | 14.3% lower | Baseline | They also buy smaller |
| Revenue per session | 10.3% higher | Baseline | Higher rate beats lower basket |
| Share of annual revenue | 1.48% of the organic figure | Baseline | Volume is the whole story |
| Year-on-year growth | 1,079% | 17% | Direction is real, base is small |
Two rows contradict the simple version of this story. According to that study, average order value was 14.3% lower on ChatGPT traffic, so the higher conversion rate is partly buying smaller baskets. Revenue per session still came out 10.3% ahead, which is the honest version of the win.
The growth figure deserves the same treatment. ChatGPT visits grew from 1,544 in January to 18,202 in December. That is a large multiple of a small number. Visibility Labs tied the first-half spike to shopping carousel features introduced in April 2025, and noted growth began flattening around August. Growth driven by one platform feature is not the same as growth driven by changing buyer behaviour, and it can flatten when the next feature does not land.
The Part Your Analytics Cannot See
This is where the study becomes more useful than its headline.
Many people ask ChatGPT for a recommendation, then search the brand name on Google and buy there. That conversion is recorded as branded organic search. The AI conversation that caused it leaves no trace in your referral report.
So the measurable AI traffic is the smaller portion of AI influence, and the gap between them is not something you can close with better tagging. Referral data undercounts by design here.
The practical response is not a new dashboard. It is a question on your post-purchase or enquiry form asking how the customer first heard about you, with an open text field. Self-reported attribution is imprecise and people forget, but it is the only instrument that catches an influence path which never touched your site.
If you sell services rather than products, this matters more, not less. A considered B2B purchase involves more research steps, and more of those steps happen somewhere you cannot log.
Size It On Your Own Numbers First
Before anyone proposes a budget shift, run this on your own analytics. It takes about fifteen minutes.
Open GA4 and build a report on session source. Filter for the AI referrers you care about: chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, copilot.microsoft.com. Compare three columns against your non-branded organic sessions: session count, conversion rate, and revenue or enquiries.
Then apply a decision rule. These bands are our working thresholds, not an industry standard, and you should adjust them to your own margin. If AI referral sessions come in under 2% of your organic sessions, treat this as a monitoring exercise and nothing more, then set a quarterly reminder and move on. If your share lands between 2% and 10%, give it a named owner and a place in the monthly report. If they exceed 10%, treat it as a channel and give it a budget line.
Most Indian SME websites we would expect to sit in the first band today. That is not a reason to ignore AI search. It is a reason to fund it from your existing content and technical work rather than from a channel that is currently producing enquiries.
The tradeoff worth naming plainly: the work that improves AI visibility is mostly the work that improves organic visibility. Clear entity information, pages that answer a question directly, accurate business facts, credible third-party mentions. You are rarely choosing between the two. You are choosing whether to pay a premium for a separate AI programme while the overlap is still this large.
One warning. Do not use a conversion rate comparison to justify cutting paid search or organic content. The base rates are not comparable in the way the headline implies, and the study measured ecommerce brands at seven and eight figures. A local services business in Mumbai has a different traffic mix entirely.
Your five-minute action today: check whether your enquiry form asks how someone found you. If it does not, add the field this week. The data only starts accumulating from the day you ask.
Tips For Measuring AI Traffic Honestly
- Build the AI referral segment once, in GA4, and save it. Rebuilding it ad hoc each quarter produces numbers that are not comparable.
- Track AI referrals as a share of organic sessions, not as a raw count. A raw count rises with total traffic and tells you nothing.
- Record the date you created the segment. Comparisons before that date are not available and people forget this three months later.
- Keep branded and non-branded organic separate in the same view. A rise in branded search alongside flat AI referrals is the attribution gap showing itself.
- Add a free-text source question to enquiry and checkout forms, and read the answers monthly rather than charting them. The value is in the phrasing, not the count.
- Do not report AI referral conversion rate on fewer than about 30 conversions in the period. Below that, the rate moves on noise.
What To Keep An Eye On
Watch the ratio between AI referral sessions and non-branded organic sessions over quarters, not months. Watch branded search volume in Search Console, because a rising brand line with flat referrals is the clearest sign that AI is influencing buyers who reach you by another route. Watch whether your own category questions produce your name when you ask an assistant directly. And watch for the study behind any statistic someone quotes at you. This one measured ecommerce, in one market, over 2025. Ask what yours would measure.
FAQs
Does a higher conversion rate mean I should invest more in AI search?
Not on its own. Conversion rate tells you about the quality of the visitors arriving. Volume tells you what the channel is worth. According to the Visibility Labs study, the volume was 1.48% of organic revenue, which makes it a signal to monitor rather than a line to fund.
Why does my GA4 show almost no ChatGPT traffic?
Two likely reasons. Your category may simply not be one people research through assistants yet. Or the influence is real but arrives as branded search after the conversation, which is the attribution gap the study describes.
Is this data relevant if I am not in ecommerce?
Partly. The direction of travel, that AI visitors arrive more decided, is plausible across categories. The specific rates are not transferable. A services business should treat the method as the takeaway and run the comparison on its own numbers.
Should I add tracking parameters to catch AI traffic?
You cannot add parameters to links you do not control, and assistants generate those links. Referrer-based segmentation plus a self-reported question on your forms is the realistic combination.
Where This Leaves You
The useful reading of this study is not that AI traffic is better. It is that AI traffic is different: more decided, smaller basket, far smaller volume, and substantially undercounted. Those four things together justify attention and an honest measurement setup. They do not yet justify moving money out of what is working.
Run the comparison on your own numbers before anyone runs it for you. If you would rather have that done properly, our AI search optimization work starts with exactly this baseline. Our guide to SEO, AEO and GEO explains what sits underneath the measurement, and how rankings turn into leads and sales covers the wider question of which numbers deserve a decision. For a worked example of measurement being rebuilt around revenue rather than sessions, see our financial services SEO case study.